AI consultation and automation
Most business owners have been told AI will change everything, and many have tried a chatbot or two without seeing much difference. The useful jobs tend to be narrower: drafting, summarising, finding information in documents and emails, and answering staff questions from your own files.
I look at how your business works and where the time goes, then tell you plainly where AI is worth trying. That can mean choosing the right model for the job or setting up agents that work within limits you set. It can also mean connecting AI to your business systems through APIs and MCP.
I use Claude, Gemini and ChatGPT myself, and I’ve built MCP integrations that connect AI to multiple mailboxes, UniFi networks, QuickBooks Online and Atera. Most of my AI work with clients is advice.
I’m just as ready to tell you when AI isn’t the right move. Scripts almost always beat AI for routine jobs. The difference now is that AI can write the scripts, but someone with an IT background still needs to read every one before it runs.
Before connecting AI to your data, it’s worth knowing who can already see what. A Microsoft 365 review shows you that quickly.
What it involves
Where AI fits
A look at your daily work to find the tasks where AI saves real time, and the ones where it adds risk or cost for little return.
Choosing models
Advice on which models suit which jobs, including when a reasoning model is worth the extra cost, and how each provider handles your data.
Agents and integrations
AI connected to your email, documents and business systems through APIs and MCP, with its access limited to what each task needs.
Scripting and automation
PowerShell, Linux scripts, Docker and automated workflows for repetitive jobs, where a plain script is cheaper and more predictable than AI.
Questions about AI consultation
Is our data safe if staff use AI tools?
It depends on the tool and the plan. Business versions of the main AI services generally don’t train on your data, while free consumer versions may. I’ll help you pick tools with suitable terms, set a simple policy for staff, and check that AI connected to your files can only see what it should.
What is MCP?
MCP, the Model Context Protocol, is an open standard for connecting AI assistants to other systems, such as your files, calendar, CRM or accounts software. It lets an assistant look things up or carry out tasks in those systems, within permissions you control.
Do we need Microsoft Copilot?
Not necessarily. Copilot can be useful for businesses that work heavily in Microsoft 365. But it costs extra per user, and it will surface anything a user has access to, including files overshared years ago. I’d tidy permissions first and trial it with a few staff before buying it for everyone.
Related services
IT policies and procedures
Written procedures that reduce risk and make IT predictable, from starters and leavers to incident response.
Microsoft 365 and Google Workspace
Microsoft 365 or Google Workspace set up, tidied or migrated, with sign-ins and email properly secured.
Fractional CTO and IT strategy
Senior IT leadership on a part-time basis, without the cost of a full-time IT director.
Interested in AI consultation?
